Open source

UC Berkeley Open-Sources LEANN Vector Index to Shrink RAG Footprints

UC Berkeley released the open-source vector index LEANN, which uses a pruned graph structure and recomputes embeddings at query time to shrink a 60M-chunk RAG index from 201GB to 6GB. It matches HNSW accuracy at 90% recall but incurs seconds-level search latency.

Summarized by AI from one outlet

Source

Latest

  1. Open source

    Microsoft releases TokenEconomizer plugin to cut costs

    Microsoft Research and UC Berkeley released TokenEconomizer, an open-source plugin that reduces coding agent costs by 18.4% to 30.9% on SWE-Bench. It achieves this by learning from past sessions without modifying the agent architecture.

    MarkTechPost

  2. Open source

    Xiaomi open-sources 0.8B OCR model

    Xiaomi open-sourced Xiaomi-OCR-0, a lightweight document understanding model with 0.8B parameters. Based on the Qwen3.5 foundation, it covers document parsing, key information extraction, and OCR visual question answering, with weights and code available on Hugging Face.

    IT之家

All news → · All rights belong to the original creators